List your company

Prompt Injection on SQuAD Inj

1.11ASR (Naive)Robustness via Referencing
-0.752411.818824.3936.9612Apr 29, 2025

Evaluation results

MethodLinks
Robustness via ReferencingBackbone=Llama3.1-405B-Instruct2025.041.110.220.560.780.11
Robustness via ReferencingBackbone=Llama3-70B-Instruct2025.042.221.441.220.441.22
Robustness via ReferencingBackbone=Qwen2-72B-Instruct2025.042.780.2220.780
SandwichBackbone=Llama3.1-405B-Instruct2025.048.1124.448.229.4433.22
SandwichBackbone=Qwen2-72B-Instruct2025.049.7815.787.783.7813.11
SandwichBackbone=Llama3-70B-Instruct2025.0410.1132.22848.3346.33
SpotlightBackbone=Llama3.1-405B-Instruct2025.0415.4457.8914.4477.6785.67
ReminderBackbone=Llama3.1-405B-Instruct2025.0419.3332.1119.5622.6742.78
SpotlightBackbone=Qwen2-72B-Instruct2025.0422.5632.7819.3379.2281.33
NoneBackbone=Llama3.1-405B-Instruct2025.0422.6772.6726.336080.78
InstructionalBackbone=Llama3.1-405B-Instruct2025.0423.5637.4423.8934.1142.56
SpotlightBackbone=Llama3-70B-Instruct2025.042667.8929.1197.5699.11
NoneBackbone=Qwen2-72B-Instruct2025.0435.3382.4431.5674.8991.67
InstructionalBackbone=Llama3-70B-Instruct2025.0442.3346.8942.5691.2275.44
ReminderBackbone=Qwen2-72B-Instruct2025.0442.6771.8940.113869.78
NoneBackbone=Llama3-70B-Instruct2025.0444.7891.6750.3398.2296.67
ReminderBackbone=Llama3-70B-Instruct2025.0446.7871.5646.4487.7869.44
InstructionalBackbone=Qwen2-72B-Instruct2025.0447.6770.7842.7850.5679.33